The AXIS C-Score: How Agent Credit Scoring Works and Why It Matters
Just as a credit score predicts a borrower's likelihood of repayment, the AXIS C-Score predicts an AI agent's reliability for resource-intensive tasks. Here is how it is calculated and how operators can use it.
Leonidas Esquire Williamson
Team Axis Trust
From Trust to Credit
The AXIS T-Score measures an agent's trustworthiness — its identity, compliance, and behavioural history. The C-Score (Credit Score) measures something related but distinct: an agent's capacity and reliability for resource-intensive operations.
Think of it this way. A T-Score answers the question should I trust this agent? A C-Score answers the question can this agent handle this task?
What the C-Score Measures
The C-Score ranges from 0 to 1000 and is computed from five components:
| Component | Weight | Description |
|---|---|---|
| Task Completion Rate | 30% | Percentage of assigned tasks completed successfully |
| Resource Efficiency | 20% | How efficiently the agent uses compute, tokens, and API calls |
| Latency Profile | 15% | Consistency and predictability of response times |
| Error Recovery | 20% | How gracefully the agent handles failures and retries |
| Capacity History | 15% | Track record with high-volume and high-stakes operations |
These components are designed to be complementary to the T-Score. An agent can have a high T-Score (trustworthy) but a low C-Score (unreliable under load). An agent with both high scores is a strong candidate for critical, resource-intensive workloads.
The Credit Tier System
Like the T-Score, the C-Score maps to five tiers with specific operational implications:
C1 — Unrated (0–199). No meaningful operational history. Suitable only for low-stakes, low-volume tasks with human oversight.
C2 — Limited (200–399). Some operational history but inconsistent performance. Appropriate for batch tasks with generous timeouts and fallback logic.
C3 — Standard (400–599). Reliable for routine workloads. Can be trusted with standard SLAs and moderate resource budgets.
C4 — Premium (600–799). Proven performance under load. Suitable for production workloads with tight SLAs and significant resource budgets.
C5 — Elite (800–1000). Exceptional operational track record. Appropriate for mission-critical workloads, real-time processing, and high-stakes automation.
How C-Score Differs From T-Score
The two scores are independent but correlated. An agent can achieve a high T-Score through careful, low-volume operation — always compliant, always transparent, but never tested at scale. The C-Score captures the operational dimension that the T-Score does not.
In practice, most operators use both scores together. A common policy pattern is:
if (tScore >= 600 && cScore >= 500) {
grantAccess("production-workload");
} else if (tScore >= 400 && cScore >= 300) {
grantAccess("staging-workload");
} else {
denyAccess("insufficient trust or credit");
}
if (tScore >= 600 && cScore >= 500) {
grantAccess("production-workload");
} else if (tScore >= 400 && cScore >= 300) {
grantAccess("staging-workload");
} else {
denyAccess("insufficient trust or credit");
}
This two-dimensional gating gives operators much finer control than either score alone.
Building Your Agent's C-Score
Unlike the T-Score, which is partly determined by factors outside the agent's direct control (such as third-party audits), the C-Score is almost entirely within the operator's influence. The most effective ways to build a strong C-Score are:
Complete tasks reliably. The single most important factor is task completion rate. An agent that consistently finishes what it starts — even if it takes longer than expected — will build a strong C-Score over time.
Handle errors gracefully. Agents that implement proper retry logic, exponential backoff, and graceful degradation score significantly higher on the error recovery component than agents that fail hard.
Report resource usage accurately. The AXIS platform rewards agents that accurately forecast and report their resource consumption. Agents that consistently over-request resources or under-report usage are penalised.
Build operational history gradually. The C-Score is designed to be conservative with new agents. A gradual ramp-up — starting with low-stakes tasks and progressively taking on more demanding workloads — is the most reliable path to a high C-Score.
Querying the C-Score
The C-Score is available on all verify API responses alongside the T-Score:
curl -H "Authorization: Bearer YOUR_API_KEY" \
https://api.axistrust.io/v1/verify/AUID-XXXXXXXXXX
# Response includes:
# {
# "auid": "AUID-XXXXXXXXXX",
# "tScore": 742,
# "tTier": "T4",
# "cScore": 681,
# "cTier": "C4",
# "lastUpdated": "2026-03-10T14:23:00Z"
# }
curl -H "Authorization: Bearer YOUR_API_KEY" \
https://api.axistrust.io/v1/verify/AUID-XXXXXXXXXX
# Response includes:
# {
# "auid": "AUID-XXXXXXXXXX",
# "tScore": 742,
# "tTier": "T4",
# "cScore": 681,
# "cTier": "C4",
# "lastUpdated": "2026-03-10T14:23:00Z"
# }
For a complete reference on all score components and their calculation methodology, see the AXIS Technical Specification [blocked].